drm/amdgpu: add sched sync for amdgpu job v2

this is an improvement for previous patch, the sched_sync is to store fence
that could be skipped as scheduled, when job is executed, we didn't need
pipeline_sync if all fences in sched_sync are signalled, otherwise insert
pipeline_sync still.

v2: handle error when adding fence to sync failed.

Signed-off-by: Chunming Zhou <David1.Zhou@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com> (v1)
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -1124,6 +1124,7 @@ struct amdgpu_job {
 	struct amdgpu_vm	*vm;
 	struct amdgpu_ring	*ring;
 	struct amdgpu_sync	sync;
+	struct amdgpu_sync	sched_sync;
 	struct amdgpu_ib	*ibs;
 	struct dma_fence	*fence; /* the hw fence */
 	uint32_t		preamble_status;

drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c

@@ -121,6 +121,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib *ib = &ibs[0];
+	struct dma_fence *tmp;
 	bool skip_preamble, need_ctx_switch;
 	unsigned patch_offset = ~0;
 	struct amdgpu_vm *vm;
@@ -160,8 +161,13 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
 		return r;
 	}
-	if (ring->funcs->emit_pipeline_sync && job && job->need_pipeline_sync)
+
+	if (ring->funcs->emit_pipeline_sync && job &&
+	    (tmp = amdgpu_sync_get_fence(&job->sched_sync))) {
+		job->need_pipeline_sync = true;
 		amdgpu_ring_emit_pipeline_sync(ring);
+		dma_fence_put(tmp);
+	}
 
 	if (vm) {
 		r = amdgpu_vm_flush(ring, job);

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

@@ -60,6 +60,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
 	(*job)->need_pipeline_sync = false;
 
 	amdgpu_sync_create(&(*job)->sync);
+	amdgpu_sync_create(&(*job)->sched_sync);
 
 	return 0;
 }
@@ -98,6 +99,7 @@ static void amdgpu_job_free_cb(struct amd_sched_job *s_job)
 
 	dma_fence_put(job->fence);
 	amdgpu_sync_free(&job->sync);
+	amdgpu_sync_free(&job->sched_sync);
 	kfree(job);
 }
 
@@ -107,6 +109,7 @@ void amdgpu_job_free(struct amdgpu_job *job)
 
 	dma_fence_put(job->fence);
 	amdgpu_sync_free(&job->sync);
+	amdgpu_sync_free(&job->sched_sync);
 	kfree(job);
 }
 
@@ -139,10 +142,10 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job)
 	struct amdgpu_vm *vm = job->vm;
 
 	struct dma_fence *fence = amdgpu_sync_get_fence(&job->sync);
+	int r;
 
 	while (fence == NULL && vm && !job->vm_id) {
 		struct amdgpu_ring *ring = job->ring;
-		int r;
 
 		r = amdgpu_vm_grab_id(vm, ring, &job->sync,
 				      &job->base.s_fence->finished,
@@ -153,9 +156,11 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job)
 		fence = amdgpu_sync_get_fence(&job->sync);
 	}
 
-	if (amd_sched_dependency_optimized(fence, sched_job->s_entity))
-		job->need_pipeline_sync = true;
-
+	if (amd_sched_dependency_optimized(fence, sched_job->s_entity)) {
+		r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence);
+		if (r)
+			DRM_ERROR("Error adding fence to sync (%d)\n", r);
+	}
 	return fence;
 }